From: James Ventre (messageboard@ventrefamily.com)
Date: Fri Aug 26 2005 - 10:57:31 GMT-3
There is a distinct difference between pinging THROUGH a device than
pinging TO a device. Pinging TO a device means he can't forward the
packets in hardware (ASIC level) - the switch has to send this up to
the CPU so it can answer hear/answer it - this means you're more at the
mercy of a busy CPU where normal traffic flows aren't necessarily
impacted (normally forwarded at the ASIC). A common example is in CatOS:
2005 Aug 01 09:12:00 EDT -04:00 %IP-6-UDP_SOCKOVFL:UDP socket overflow
from Source IP: 10.10.10, Destination port:
161
2005 Aug 01 09:12:00 EDT -04:00 %IP-6-UDP_SOCKOVFL:UDP socket overflow
from Source IP: 10.10.10, Destination port:
161
There was just too much SNMP activity to send to the CPU for processing
(so it dropped a bunch of packets) - but packets that are traveling
through this switch (server traffic) are not impacted.
James
csc david wrote:
>Hi, there are 6509,3550,3524,2924 in my network, the management vlan is Vlan 1, and I config each switch with an ip address. When I ping the 3524,2924 interface vlan 1 ip address, it will reply 1ms,4ms or even to 16ms sometime, the cpu load on 3524 and 2924 is normal. but when I ping the PCs linked on these 3524,2924, it is < 1 ms, and I ping the 6509,3550, is also <1 ms. The 6509,3550,3524 all links with Gigabit, so I could not understand why ping the 3524,2924 is so slow?
>Can anyone help me? thanks.
This archive was generated by hypermail 2.1.4 : Sun Sep 04 2005 - 17:01:20 GMT-3